CD70 Inversely Regulates Regulatory T Cells and Invariant NKT Cells and Modulates Type 1 Diabetes in NOD Mice

In conjunction with Ag recognition by the TCR, T cell activation is carefully regulated by various receptor/ligand molecules on the surface of responder T cells and APC. This second signal, referred to as costimulation, enables full T cell activation, proliferation, and differentiation. Early T cell priming events require CD28 costimulatory molecule interactions with B7 ligands (1). However, later phases of the immune response rely on differing combinatorial signals drawn from a collection of TNF superfamily receptor (TNFR)/TNF superfamily members (including CD27–CD70, CD30/CD30L, OX40/OX40L, CD137/CD137L, HVEM, and GITR/GITRL), which vary considerably in location and timing of expression (2–4). Costimulatory molecule interactions also contribute to thymocyte selection, which is controlled by the intensity with which MHC-peptide complexes on hematopoietic and epithelial cells induce TCR signaling in CD4/CD8 double-positive (DP) thymocytes (5). Currently, most data support that costimulatory molecule interactions alter thymic selection by modifying the TCR signaling threshold (5). This is especially evident for immunoregulatory T cell populations such as invariant NKT (iNKT) cells and regulatory T cells (Tregs) that seemingly require higher levels of signaling for their development (6, 7).

CD27 is a transmembrane homodimer with subunits of 55 kDa that is expressed on B cells, NK cells, and T cells (8–10). CD27 is constitutively expressed at high levels on naive T cells and increases during early T cell activation (11, 12). Expression is also high on central memory T cells, whereas late-stage effector and effector memory T cells downregulate CD27 (13). CD27 signaling is tightly controlled by the very limited expression of its ligand CD70 that is constitutively expressed by discrete cellular niches within the thymus and gut and is transiently expressed by mature dendritic cells (DCs), macrophages, activated NK cells, T cells, and B cells (14, 15). CD27–CD70 engagement appears to be important for T cell expansion, survival, and long-term memory formation, which manifests in the increased susceptibility of CD27- and CD70-deficient mice to numerous viruses and cancer (16–24).

Several reports show that CD27–CD70 costimulation is involved in autoimmunity, where signaling through this pathway is usually associated with increased pathology. CD70 is overexpressed on pathogenic CD4⁺ T cells from patients suffering from rheumatoid arthritis and systemic lupus erythematosus as well as MRL/lpr mice that develop a lupus-like disease (25, 26). Furthermore, CD70 blockade leads to the amelioration of pathology in mouse models of arthritis and various inflammatory bowel diseases (27, 28). By contrast, CD27–CD70 interactions play a protective role in experimental autoimmune encephalomyelitis, a mouse model of multiple sclerosis, by impairing Th17 effector cells that are critical for autoimmunity (29). Thus, CD27–CD70 interactions can have diametric effects on autoimmune pathologies, which probably depends on the role of various T cell subsets in the pathogenesis of these different disorders.

In this study, we asked whether the CD27–CD70 pathway contributes to autoimmune type 1 diabetes (T1D) in the NOD mouse model given that autoreactive T cells play a central role in disease development by destroying pancreatic β cells. Disrupting the Cd70 gene accelerated T1D development in NOD mice, indicating CD27–CD70 costimulation elicits immunological tolerance induction responses inhibiting disease onset. We show that CD27–CD70 interactions limit circulating T cell levels and modulate thymocyte selection by controlling the ratio of DP to double-negative (DN) thymocytes. Notably, CD70 ablation diametrically skewed the development of iNKT cells and Foxp3⁺Helios⁺ Tregs, which is significant because T1D development in NOD mice is partly due to numerical and functional deficiencies in both cell types (30–39). Indeed, NOD mice are protected from T1D by treatment with the iNKT cell agonist α-galactosylceramide (α-GalCer) (36–38, 40) or adoptively transferred iNKT cells (41, 42), and CD1d-deficient mice that lack iNKT cells develop accelerated T1D (36, 39). Moreover, compared with other mouse strains, NOD Tregs have a reduced capacity to suppress conventional T cells (33, 43, 44) partly because of a hypomorphic IL-2 allelic variant and additional susceptibility genes (45). Furthermore, it has been reported that NOD Treg functionality declines with age (33, 34), although later studies have disputed this finding (46, 47). Because CD27–CD70 signaling was found to regulate both immunoregulatory and conventional T lymphocyte subsets involved in β cell autoimmunity, the aim of the current study was to test the hypothesis that alterations in one or more T cell populations accounts for CD70-mediated modulation of T1D.

CD70-deficient NOD/ShiLtDvs (Cd70^−/−) mice were generated at The Jackson Laboratory using transcription activator-like effector nucleases (TALENs) designed to disrupt the Cd70 start site. The TALENs were designed using the online software Zinc Finger Targeter (48, 49) and were constructed using the fast ligation-based automatable solid-phase high-throughput assembly system (50). These were designed to target the cleavage domain and a DNA binding domain in exon 1. The left TALEN targets the 18-bp sequence 5′-TGTTTATTACTGTGTTTT-3′, whereas the right TALEN targets the 18-bp sequence 5′-TAGCGGACTACTCAGTAA-3′. Following binding and dimerization, the TALEN pair should cause a double-stranded genomic break within the 16 bases of intervening sequence (5′-GCTGTTGGTTTCATTG-3′). The two TALENs were prepared as mRNA and microinjected directly into NOD/ShiLtJ zygotes at a concentration of 50 μg/μl each. A total of 304 embryos were transferred to pseudopregnant females, ultimately yielding 99 offspring. DNA was extracted from tail tips and screened for TALEN-induced mutations by PCR amplification using forward (5′-CAACAACGTGTGCGTCAATC-3′) and reverse (5′-CAGCCCAATACTGAACCTCTAAT-3′) primers. Sanger sequencing of the PCR products identified a male mutant mouse with a 363-bp deletion at the start site of exon 1 of Cd70. This founder was backcrossed to NOD/ShiLtJ females, and the heterozygous progeny intercrossed to generate littermates of all three genotypes. Genotyping was performed by PCR using the primers described above (wild-type [WT] allele = 785 bp, mutant allele = 422 bp). This strain, NOD/ShiLtJ-Cd70<em1Mvw>/Mvw (stock no. 24167), was imported from The Jackson Laboratory and maintained in a specific pathogen-free research colony. A previously described NOD stock carrying an inactivated Rag1 gene [NOD.129S7(B6)-Rag1<tm1Mom>/SzJ, designated NOD.Rag1^−/−] (51) was used to create NOD.Rag1^−/−.Cd70^−/− mice (NOD.Cg-Rag1<tm1Mom> Cd70<em1Mvw>/Jpd). Previously described NOD stocks respectively transgenically expressing the TCR from the diabetogenic BDC2.5 CD4⁺ T cell clone [NOD.Cg-Tg(TcraBDC2.5,TcrbBDC2.5)1Doi/DoiJ] (52) and expressing GFP off the mouse Foxp3 promoter [NOD/ShiLt-Tg(Foxp3-EGFP/cre)1cJbs/J] (53) were intercrossed with Cd70^−/− mice to generate NOD.BDC2.5.Foxp3^eGFP.Cd70^−/− mice. Previously described NOD.Foxp3-GFP-Cre.Rosa26-YFP mice (54) were obtained as a gift from the laboratory of Jeffrey A. Bluestone.

Diabetes was monitored weekly by measuring glycosuria with Bayer Diastix (Bayer, Diagnostics Division, Elkhart, IN), with disease onset defined by values ≥3 on two consecutive days. Insulitis was assessed as previously described (55). Briefly, pancreata from indicated mice were fixed in a 10% neutral buffered formalin solution and embedded with paraffin using automatic processors (VIP6; Leica) with graded levels of ethanol (70–100%) and xylene, after which samples were stained with H&E. Insulitis scores were assigned according to a previously designated classification ranging from 0 (normal β cell mass) to 4 (complete destruction) (56).

Preparation of islet infiltrating leukocytes was adapted from a previous publication (57). Pancreata from 7-wk-old NOD.Rag1^−/− and NOD.Rag1^−/−.Cd70^−/− splenocyte or T cell recipients were inflated with a Liberase Thermolysin Low solution (0.91 U/ml; Roche Diagnostics) in HBSS via the bile duct. The inflated pancreata were removed and incubated at 37°C for 14 min. After digestion, pancreata were washed three times with HBSS, resuspended in 5 ml of HBSS, and poured into 60 × 15–mm glass petri dishes. Islets from every mouse were hand-picked into a 1.5-ml Eppendorf tube under a dissecting microscope. After centrifugation, islets were resuspended in 750 μl of enzyme-free Cell Dissociation Buffer (Invitrogen) to obtain single-cell suspensions. Cells were then washed once with HBSS, resuspended in FACS buffer, and stained with the indicated fluorochrome-conjugated Abs.

Ab staining and flow cytometry analysis protocols have been previously described (35). Whole spleen, thymus, and pancreatic lymph nodes (pLNs) from euthanized mice were dispersed into single-cell suspensions using glass homogenizers after which erythrocytes were lysed using an ammonium chloride–based lysis buffer, and cells were counted using a BD Accuri C6 flow cytometer. Cell suspensions were Fc receptor blocked using anti-CD16/CD32 (2.4G2) from BioXCell (West Lebanon, NH). Abs purchased from BD Bioscience and eBioscience were used to stain the following surface molecules: CD8 (clone; 53–6.7), CD4 (RM4-5), TCRβ (H57-597), Ly-6G (RB6-8C5), CD70 (FR70), isotype control rat IgG2b (A95-1), CD45R (RA3-6B2), CD11b (M1/70), CD25 (PC61), CD86 (GL1), CD1d (1B1), CD27 (LG.7F9), pCDA1 (JF05-1C2.4.1), and T cell Ig and ITIM domain (TIGIT) (GIGD7). iNKT cells were identified using PBS57-loaded and control CD1d tetramers supplied by the National Institutes of Health Tetramer Core Facility. iNKT cell cytokine staining was performed as previously described (58). Briefly, splenocytes from mice injected i.v. 2 h previously with α-GalCer (4 μg per mouse) or DMSO were stained with anti–IFN-γ (XMG1.2) and anti–IL-4 (11B11) after cells were surface stained with CD1d tetramer, anti-CD4, and anti-TCRβ and fixed and permeabilized with solutions from the BD Biosciences Cytofix/Cytoperm kit. Intranuclear staining was performed using the Transcription Factor Staining Buffer set from eBioscience and Abs against Foxp3 (FJK-16s) and Helios (22F6) for Tregs and PLZF (R17-809; BD Biosciences) T-bet (O4-46; BD Biosciences), and RORγt (B2D; eBiosciences) for iNKT cells. Stained cells were washed once with FACS buffer and acquired using an LSR Fortessa or BD Accuri C6 flow cytometer (BD Biosciences). Fluorescence minus one controls were used to identify positive and negative populations (Supplemental Figs. 1, 2). Propidium iodide or a fixable viability dye from BD Bioscience was used for exclusion of dead cells during analysis. All flow data were analyzed using FlowJo software (version 10.5.0; Treestar, Palo Alto, CA).

To validate the ablation of CD70 protein expression in Cd70^−/− mice, 2 × 10⁶ splenocytes from Cd70^−/− or standard NOD mice were seeded in RPMI 1640 (containing 10% FBS, 1% penicillin–streptomycin–neomycin, and 1% 2-ME) with or without 1 μg/ml LPS and cultured at 37°C for 24 h. Thereafter, cells were stained with Abs against CD11c, CD86, CD70, and an isotype control for flow cytometric analysis. A previously described protocol (59) was used to compare the ability of NOD and Cd70^−/− CD25⁺CD4⁺ Tregs to suppress CD25⁻CD4⁺ effector T cell proliferation. Different concentrations of Tregs were cultured with 1 μg/ml anti-CD3 (145-2C11), 5 × 10⁴ CFSE-labeled NOD effector T cells, and 1 × 10⁴ CD11c⁺ DCs purified from Cd70^−/− mice. Tregs, effector T cells, and DCs were purified using magnetic beads (Miltenyi Biotec) according to the manufacturer’s directions. Proliferation was measured after 3 d by flow cytometry.

NOD.Rag1^−/− mice were injected i.v. with 5 × 10⁶ total splenocytes from 7- to 10-wk-old NOD and Cd70^−/− mice. Blood samples were analyzed by flow cytometry weekly until recipient mice were euthanized at 3 wk after injection. In another study, NOD.Rag1^−/− or NOD.Rag1^−/−.Cd70^−/− mice were adoptively transferred with 5 × 10⁶ total T cells purified from splenocytes of 6-wk-old NOD mice by magnetic bead separation. One group of recipients was monitored for T1D development until 22 wk posttransfer, while a second cohort was euthanized at 7 wk posttransfer to analyze Tregs in the spleen, pLNs and pancreatic islets. In a separate experiment, NOD.Rag1^−/− and NOD.Rag1^−/−.Cd70^−/− mice were i.v. injected with 4.5 × 10⁶ Foxp3-negative T cells that were sorted from NOD.Foxp3-GFP-Cre.Rosa26-YFP donors using a BD FACS Aria III. Recipient mice were euthanized after the development of T1D or ulcerative colitis. Another experiment injected NOD.Rag1^−/− mice with 2 × 10⁶ splenocytes from diabetic NOD mice admixed with CD25⁺CD4⁺ Tregs purified from Cd70^−/− or standard NOD splenocytes. An aliquot of each Treg preparation was stained for Foxp3 before transfer. Treg concentrations were then adjusted so that each NOD.Rag1^−/− recipient received 2 × 10⁵ Foxp3⁺CD25⁺CD4⁺ cells.

Bone marrow (BM) cells were collected from tibias and femurs of Cd70^−/− and NOD mice. T cells were removed using CD4 and CD8 Ab conjugated microbeads (Miltenyi Biotec). Cd70^−/− and NOD mice were lethally irradiated (6 Gy 2 h apart) and injected i.v. with 5 × 10⁶ purified BM cells from Cd70^−/− and NOD donors in a crisscross design. After 8 wk, thymi, spleens, and pLNs were collected and analyzed by flow cytometry.

Starting at 5–6 wk of age, female Cd70^−/− mice were i.p. injected twice a week with PBS or 50 μg of anti-CD27 (AT-124) provided by Celldex Therapeutics Hampton, NJ (60). Standard NOD mice were treated with PBS as a control. Mice were treated for 4 wk, after which their spleens and thymi were removed for analysis by flow cytometry.

Wilcoxon matched-pairs signed rank test, Mann–Whitney U test, or two-way ANOVA was used for comparisons among groups as indicated. Log-rank test was used for the comparison of T1D incidence. All statistical analyses were performed using Prism, version 8.0 (GraphPad Software, La Jolla, CA).

A pair of TALENs designed to target the Cd70 gene transcriptional start site was delivered by pronuclear injection into NOD zygotes. TALEN off-target mutations were avoided by choosing a unique target sequence that differed from any other site in the genome (61). Ninety-nine offspring were obtained and analyzed for mutations in the Cd70 gene. Of these, four mice were identified with deletions capable of disrupting CD70 expression, including a male with a 363-bp deletion that was selected as the founder for the CD70 knockout line. This mouse was backcrossed to standard NOD females to verify germline transmission of the mutant allele (Fig. 1A), generating the line NOD/ShiLtJ-Cd70<em1Mvw>/Mvw (stock no. 24167). Subsequent generations from this breeding scheme produced a Mendelian inheritance pattern of mutant and WT alleles, indicating that the mutation was introduced into the embryo at the one cell stage. CD70 protein ablation was confirmed by flow cytometry using LPS-stimulated splenic DCs isolated from homozygous mutant and WT NOD mice (Fig. 1B). We also confirmed that CD70 deletion does not affect the overall ability of APCs to mature, as mutant DCs retained their ability to upregulate the T cell costimulatory molecule CD86 in response to LPS stimulation (Fig. 1C, 1D).

Male and female NOD mice homozygous and heterozygous for mutant and WT Cd70 alleles were monitored for the development of insulitis and T1D. Cd70^−/− mice developed more-severe insulitis and accelerated diabetes compared with WT littermates (Fig. 2). Diabetes in heterozygous mice was intermediate between WT and Cd70^−/− mice, suggesting that disease is partially suppressed by one copy of the WT Cd70 allele. These results show that CD70 elicits immune effects that limit T1D development.

Splenocytes and thymocytes from 8- to 10-wk-old sex-matched Cd70^+/+, Cd70^+/−, and Cd70^−/− mice were analyzed by flow cytometry to determine the effect of CD70 ablation on immune cell populations, including subsets that contribute to T1D pathogenesis. No differences were found in the frequency or number of total lymphocytes, DCs, and granulocytes (Table I). However, both the frequency and number of splenic αβ T lymphocytes were increased in Cd70^−/− mice compared with Cd70^+/− and Cd70^+/+ mice (Fig. 3A, 3B). The ratio of CD4 and CD8 T cells was similar among mice of the different genotypes (data not shown). Cd70^−/− mice had a lower frequency of splenic B cells compared with Cd70^+/+ and Cd70^+/− mice; however, the number of B cells was similar among the different genotypes (Fig. 3A, 3B). CD70 ablation affected thymocyte development by respectively increasing and decreasing the frequency of the DN and DP subsets compared with CD70-intact NOD mice (Fig. 3C–E). Numerically, DP and CD4 single-positive thymocytes were lower in CD70-deficient versus -intact mice. These collective findings indicate that CD70 regulates thymic T cell development and peripheral T cell levels.

To establish whether CD70 ablation affects CD4⁺Foxp3⁺ Tregs that are important for suppressing diabetogenic T cell responses, CD70-deficient and -intact NOD mice were compared for Treg levels. The frequency of Tregs was significantly reduced in thymus, spleen, pLNs, and blood of CD70-deficient mice (Fig. 4A). Furthermore, Tregs were numerically reduced in the thymus and spleen of CD70-deficient compared with -intact mice (Fig. 4B). Splenic Tregs from Cd70^−/− mice expressed less Foxp3 compared with CD70-intact mice (Fig. 4C). However, they expressed more of the immune receptor TIGIT in the thymus and spleen and more CD25 in the thymus (Fig. 4D, 4E). To determine whether CD70 is involved in maintaining the stability of Tregs, thymic and splenic Tregs were costained with an Ab against the transcription factor Helios, which is expressed very early in Foxp3⁺Treg development and is required for maintaining the anergic and suppressive phenotype of these cells (62). We found that the Treg reduction in Cd70^−/− mice was due to a decrease in Helios⁺ and not Helios⁻ Tregs, suggesting that Treg stability is affected by CD70 ablation (Fig. 4F, 4G). Next, we assessed if CD70 deletion impacts the development of β cell islet-specific Tregs, as Ag-specific Tregs play an important role in suppressing diabetogenic immune responses (63). Tregs were analyzed from CD70-deficient and -intact NOD mice transgenically expressing the TCR from the BDC2.5 CD4 T cell clone and also GFP off the Foxp3 promoter (NOD.BDC2.5.Foxp3^eGFP). BDC2.5 clonotypes make up >80% of T cells in NOD.BDC2.5 TCR transgenic mice (64). The frequency of Tregs was significantly reduced in the thymus, spleen, and pLN of CD70-deficient than -intact NOD.BDC2.5.Foxp3^eGFP mice (Fig. 4H). This stock also had reduced numbers of Tregs in the thymus and pLN but not the spleen (Fig. 4I), which is likely because Cd70^−/−.BDC2.5.Foxp3^eGFP mice had larger spleens than their CD70-sufficient counterparts (data not shown). Interestingly, ablation of CD70 led to a greater reduction in the frequency of Tregs in NOD.BDC2.5.Foxp3^eGFP than nontransgenic NOD mice (Fig. 4A versus Fig. 4H). These results show that CD70 is important for the development of Tregs, including clonotypes that recognize β cell Ags, and that CD70 promotes the generation of Helios⁺ Tregs. The reduced number of Tregs in Cd70^−/− mice may underlie why this strain has higher levels of circulating T cells compared with standard NOD mice (Fig. 3A, 3B). Thus, young 3-wk-old mice were analyzed to determine whether increased T cell levels develop downstream of or alongside reduced Treg levels in Cd70^−/− compared with standard NOD mice. We found that the number of total splenic T cells and Tregs was respectively similar and reduced between Cd70^−/− and NOD mice (Fig. 4J, 4K). This result suggests that the higher levels of T cells in Cd70^−/− mice are not the consequence of altered thymic development but rather from impaired peripheral homeostasis, which is probably due to fewer Tregs.

Both hematopoietic and nonhematopoietic cell types are required for intrathymic Treg generation (65). To investigate whether one or both of these compartments affects Treg development through CD70 signaling, we created chimeric mice by transferring T cell–depleted NOD or Cd70^−/− BM into lethally irradiated NOD or Cd70^−/− recipients in a crisscross experimental design. Analysis of mice 8 wk after reconstitution found that recipients of Cd70^−/− BM had lower frequencies of thymic and splenic Tregs than recipients of WT BM (Fig. 5A). However, it was only in Cd70^−/− recipients of Cd70^−/− BM that the reduction in Tregs was statistically lower than mice that received NOD BM. Tregs were numerically lower in the thymus and spleen of Cd70^−/− BM recipients compared with NOD recipients (Fig. 5B). Further analysis found that the frequency and number of Helios⁺ and Helios⁻ Tregs were lower in recipients of Cd70^−/− than NOD BM (Fig. 5C, 5D). These results indicate that CD70 signaling supports Treg development, including Helios⁺ and Helios⁻ subsets, through a hematopoietic rather than nonhematopoietic cell type(s), although nonhematopoietic cells appear to make some contribution.

Because Tregs from Cd70^−/− mice express less Foxp3 but more TIGIT and CD25 than WT Tregs (Fig. 4C–E), we tested whether CD70 and WT Tregs have different effector functions. On a per cell basis, Tregs from NOD or Cd70^−/− mice equally suppressed the activation of conventional NOD CD25⁻CD4⁺ T cells in an in vitro suppression assay that used DCs from CD70-deficient mice as APCs (Fig. 6A, 6B). To determine whether the suppressor function of Tregs is fundamentally altered when these cells develop in the absence of CD70 signaling, NOD.Rag1^−/− mice were injected with either NOD or Cd70^−/− Tregs and diabetic NOD splenocytes. T1D developed at a similar rate in both groups, demonstrating that the suppressive functions of NOD and Cd70^−/− Tregs were comparable (Fig. 6C). The frequency and number of splenic Tregs at diabetes onset were similar between groups (Fig. 6D, 6E), indicating that the experiment was not affected by changes in Treg levels after adoptive transfer. Together, these results demonstrate that Tregs isolated from Cd70^−/− and NOD mice are functionally equivalent and that thymic development in the absence of CD70 does not inherently alter the capacity of Tregs to suppress conventional T cell responses.

To determine whether accelerated T1D in Cd70^−/− mice is associated with the reduced level of Tregs in this strain, we adoptively transferred 5 × 10⁶ splenocytes from CD70-deficient or -intact mice into CD70-sufficient NOD.Rag1^−/− mice. As expected, freshly isolated splenocytes from Cd70^−/− donors had a lower frequency of Tregs than WT splenocytes (Fig. 7A). However, analysis of peripheral blood from recipient mice starting from 6 d posttransfer found that Tregs from Cd70^−/− and WT donors had expanded to the same level (Fig. 7B). Because this study was no longer useful to investigate an association between Treg levels and T1D susceptibility, all recipient mice were sacrificed at 3 wk after injection. No differences were found in the frequency or number of splenic Tregs between Cd70^−/− and WT recipients (Fig. 7C and data not shown). Our finding that Cd70^−/− Tregs are capable of increasing to WT levels when transferred into a CD70-intact environment indicates that the Treg deficit of Cd70^−/− mice is due to an absence of CD70 signaling from an extrinsic source rather than from a Treg-intrinsic defect. To establish if CD70 expressed by a nonlymphoid cell population is responsible for controlling circulating Treg levels, NOD.Rag1^−/−.Cd70^−/− and NOD.Rag1^−/− mice were analyzed 7–10 wk after injection with purified T cells from standard NOD mice. Tregs were significantly reduced in spleen, pLN, and pancreatic islet infiltrates of NOD.Rag1^−/−.Cd70^−/− mice compared with NOD.Rag1^−/− recipients (Fig. 7D). This was due to a decrease in Helios⁺ rather than Helios⁻ Tregs (Fig. 7E). Unlike Tregs, CD4⁺ and CD8⁺ T cell levels were similar between treatment groups (Fig. 7F, 7G). In a separate cohort of mice, we found that T1D was accelerated in NOD.Rag1^−/−.Cd70^−/− versus NOD.Rag1^−/− NOD T cell recipients (Fig. 7H), which was not affected by CD27 expression on Tregs (Supplemental Fig. 3A). These results show that CD70 expressed by a nonlymphoid population of cells sustains circulating Helios⁺ Treg levels and inhibits T1D development. To address the possibility that differences in T1D progression in Cd70^−/− mice and NOD mice were due to differential Treg levels, we asked whether in the absence of Tregs the two strains would develop a similar rate of disease. NOD.Rag1^−/−. Cd70^−/− and NOD.Rag1^−/− mice were adoptively transferred with T cells from NOD.Foxp3-GFP-Cre.Rosa26-YFP mice after flow cytometric depletion of Foxp3⁺Tregs and cells that had at some point expressed Foxp3. We found that without these cells, Cd70^−/− mice did not develop accelerated T1D compared with WT mice (Fig. 7I). Indeed, whereas all NOD.Rag1^−/− mice succumbed to T1D by 4 wk posttransfer, two out of four NOD.Rag1^−/−.Cd70^−/− recipients remained T1D free until 7 wk posttransfer when they developed ulcerative colitis instead. This result indicates that Tregs are required for a nonlymphoid population of cells to suppress T1D through CD70 signaling.

Invariant NKT cells are another immunoregulatory T cell population that modulates T1D development in the NOD mouse. Because T cell costimulatory molecules regulate TCR signaling, which directs lineage determining factors such as Foxp3 for Tregs, we asked whether CD27–CD70 interactions affect the development of iNKT cells. iNKT cells were quantified in Cd70^+/+, Cd70^+/−, and Cd70^−/− littermates using a mouse PBS57-loaded CD1d tetramer. Thymic and splenic iNKT cells were proportionally and numerically higher in Cd70^−/− than Cd70^+/+ mice, whereas iNKT cell levels in Cd70^+/− mice were respectively similar to Cd70^+/+ and Cd70^−/− mice in the thymus and spleen (Fig. 8A, 8B). Previous reports showed an inverse relationship between CD1d expression on mouse DP thymocytes and the frequency of thymic iNKT cells (66–68). Therefore, we analyzed littermates of all three genotypes for CD1d expression. We observed that CD70 expression was correlated to CD1d expression (Fig. 8C), which subsequent reciprocal BM transfer experiments identified is due to CD70 signaling from a nonlymphoid source more than the lymphoid compartment (Fig. 8D). Next, we assessed how CD70 ablation affects the development of thymic iNKT cell lineages that segregate into iNKT1, iNKT2, and iNKT17 subsets according to the differential expression of transcription factors T-bet, PLZF, and RORγt (69). The absence of CD70 respectively increased and decreased the frequency of iNKT1 and iNKT2 cells (Fig. 8E–G) so that the iNKT1/iNKT2 ratio was ∼8:1 in Cd70^−/− mice compared with ∼3:1 in WT mice. CD70 deficiency had no effect on the frequency of iNKT17 cells, nor did it alter the ratio of CD4⁺ and CD8⁻CD4⁻ DN iNKT cells in central and peripheral lymphoid organs (data not shown). To test if CD70 deficiency alters iNKT cell function, splenic iNKT cells from NOD and Cd70^−/− mice were measured for IFN-γ and IL-4 production 2 h after injection with the iNKT cell superagonist α-GalCer. A higher frequency and number of IFN-γ and IL-4 positive iNKT cells was detected in Cd70^−/− mice (Fig. 8H, 8I). We also found that Cd70^−/− iNKT cells produced more cytokines than NOD iNKT cells on a per cell basis, (Supplemental Fig. 1E). These results suggest that CD70 signaling decreases the fraction of iNKT cells that become activated after stimulation while also limiting the amount of cytokines they produce. Because Tregs have been reported to negatively regulate iNKT cell development (70), we measured iNKT cells in standard NOD mice after Treg depletion using an anti-CD25 Ab. iNKT cell levels did not change despite a large decrease in Tregs (Supplemental Fig. 4), which suggests that reduced Treg levels is not the underlying reason why Cd70^−/− mice produce more iNKT cells. In summary, these data indicate that CD70 inhibits iNKT cell development, limits the differentiation of iNKT1 cells in favor of iNKT2 cells, and reduces iNKT cell effector functions after activation by a strong cognate ligand.

Most of the effects of CD70 have been attributed to intracellular signaling events induced by CD27. However, experimental evidence exists that CD70 ligation can evoke reverse signaling cascades that regulate the expansion and differentiation of immune cells (20, 71). To determine whether the altered iNKT cell and Treg phenotypes in Cd70^−/− mice are due to a lack of CD70 or CD27 signaling, we injected Cd70^−/− mice for 4 wk with a CD27 agonist Ab (clone AT-124). The treatment downregulated CD27 expression on Tregs, CD4⁺ T cells, and CD8⁺ T cells, which was probably a response to chronic CD27 stimulation (Supplemental Fig. 3B). Ab treatment increased the frequency, but not the number, of thymic Tregs compared with control mice (Fig. 9A, 9B). We observed a similar but more-striking effect in the spleen where Ab treatment increased Treg numbers beyond those found in NOD mice. Analysis of Treg subsets found that anti-CD27 treatment respectively increased and did not affect the frequency of Helios⁺ and Helios⁻ Tregs (Fig. 9C, 9D). In contrast to Tregs, anti-CD27 Ab treatment decreased the number of thymic iNKT cells, whereas splenic iNKT cell frequency and numbers were reduced below standard NOD mouse levels (Fig. 9E, 9F). We also found that anti-CD27 treatment increased CD1d expression compared with PBS-treated Cd70^−/− mice but only in the spleen (Fig. 9G). To determine whether the immune responses elicited by anti-CD27 Ab administration would reduce islet inflammation in Cd70^−/− mice, pancreata from the different experimental groups were examined for insulitis levels after 4 wk of treatment. Insulitis was significantly lower in Cd70^−/− mice treated with anti-CD27 compared with PBS (Fig. 9H). However, anti-CD27 Ab administration failed to suppress the onset of T1D in either CD70-deficient or -intact mice when injections were continued until 30 wk of age (data not shown). This is likely because AT-124 eventually lost its effect on Tregs that were found at similar levels between anti-CD27 Ab and control treated mice at the end of the study. Overall, these results support that the altered levels of immunoregulatory cell populations and islet inflammation observed in Cd70^−/− mice are due to an absence of CD27 intercellular signaling rather than from a lack of reverse signaling from CD70 ligation.

Previous studies that examined the contribution of the CD27–CD70 costimulation axis to autoimmunity have found that this pathway often plays Janus-like opposing roles in different disorders (25, 27–29, 72). This is likely because individual CD27-expressing cell types, such as T cells, make distinct contributions to these various diseases. The current study is the first, to our knowledge, to describe the contribution of the CD27–CD70 axis for T1D. This disease develops primarily from defects in T cell tolerance mechanisms. These defects include the aberrant apoptosis-mediated deletion of autoreactive T cells (73, 74) and the impaired ability of immunoregulatory populations, such as Tregs and iNKT cells, to limit the development and/or activation of pathogenic T lymphocytes (30–34). Although CD27 signaling modulates both conventional and immunoregulatory T cells, our finding that CD70 ablation exacerbates T1D indicates that the combined effects of CD27–CD70 costimulation favors tolerogenic over pathogenic immune responses in this disease.

We focused on how CD70 regulates T lymphocyte subsets because we found that the T cell compartment was significantly altered in Cd70^−/− mice. Our observation that CD70 ablation reduced the number of DP thymocytes is consistent with a previous report that CD27 mAb blockade in 129Sv mixed background mice impedes pre–T cell expansion and differentiation, which resulted in a specific decrease in the absolute number of DP thymocytes without affecting the size of the DN compartment (75). In contrast, studies using B6 mice lacking either CD27 or CD70 found normal steady-state immune compartments including thymocyte subsets (22, 76). This discrepancy might be due to the genetic background of the mice employed as several negative selection-associated and prosurvival genes are reported to be improperly regulated during negative selection in the NOD mouse (77–80). The altered thymic selection in Cd70^−/− mice did not underlie why peripheral T cell levels are increased in this strain. Rather, this phenomenon is probably because CD70 ablation impairs the development of Tregs that usually control peripheral T cell expansion. These higher circulating T cell levels may contribute to why T1D onset is more aggressive in Cd70^−/− mice.

Although several studies have shown that the CD27–CD70 pathway is involved in the development and/or the induction of Tregs (21, 76, 81, 82), no evidence has been reported that the reduced levels of Tregs in CD27- and CD70-deficient mice renders these stocks more susceptible to spontaneous autoimmunity. This is likely because knockout mice were generated on the B6 genetic background, which is a nonautoimmune prone strain. In contrast, the NOD mouse that spontaneously develops autoimmunity provides a useful model to address this question, particularly as NOD Tregs are already impaired due to a hypomorphic Il2 allelic variant (83) and additional susceptibility genes (46). Overall, Treg levels were reduced by ∼20% in Cd70^−/− mice compared with controls, which is similar to the reduction in CD27- and CD70-deficient B6 mice. An important mechanism through which CD27–CD70 signaling is reported to increase Treg levels is by augmenting IL-2 production by Foxp3-negative CD4⁺ T cells (82). Thus, it is somewhat surprising that overall Treg levels were not more reduced on the NOD background that produces less IL-2 than B6 mice (83, 84). Our study shows that CD27 costimulation supports the generation of β cell–specific Tregs that exert their suppressive function within the islets of Langerhans, which is vital for controlling pancreatic inflammation (63). BDC2.5 TCR Tregs were more dependent on CD27 costimulation compared with polyclonal Tregs in standard NOD mice. This could be due to the affinity with which the BDC2.5 TCR binds self-antigen MHC complexes during thymic selection, as Croquet et al. (76) have shown that CD27 costimulation is more important for the development of high- than low-affinity Tregs.

CD70 expression in the thymus is mostly restricted to Aire⁺ medullary thymic epithelial cells (mTECs) and a small number of DCs (76). Our reciprocal BM transfers found that CD70 on hematopoietic cell types rather than mTECs is more important for promoting Treg development in NOD mice. This differs from a previous report showing that mTECs and DCs jointly contribute to Treg development in B6 background mice (76). The discrepancy may be due to variation in positive and negative selection between B6 and NOD mice, including that NOD mice have comparatively high numbers of DP thymocytes competing for limited selection niches, including mTECs (85). We found that CD27–CD70 costimulation is respectively supportive and dispensable for the generation of Helios⁺ and Helios⁻ Tregs. Although the presence or absence of Helios expression does not unambiguously distinguish Tregs of thymic and peripheral origin (62, 86, 87), there is general consensus that immediate precursors of thymic-derived Tregs have high expression of Helios and that this molecule identifies more thymic-derived Tregs than any other single biomarker (87, 88). Therefore, one explanation of our findings is that CD27–CD70 costimulation enhances the selection of thymic-derived Tregs. Another possibility is that CD70 signaling improves Treg stability after thymic selection as Helios expression is associated with enhanced Treg differentiation and survival (89, 90). Our contrary finding that CD70 deficiency reduces both Helios⁻ and Helios⁺ Tregs in BM chimeras is likely an artifact of the BM reconstitution procedure as thymic selection of Tregs in lethally irradiated BM recipients probably differs from normal neonatal Treg selection in unmanipulated mice.

We found no difference on a per cell basis in the suppressive function of Tregs from CD70-intact and -deficient mice, even though Tregs from the latter strain respectively expressed less Foxp3 and more TIGIT, which are both markers of Treg effector function (91, 92). In our in vivo suppression assay, it is possible that Tregs from CD70-deficient mice were initially impaired but regained their suppressor functions when transferred into a CD70-intact environment. However, CD70 most likely regulates T1D development by affecting the number rather than the function of Tregs as a previous report has shown that CD27-deficient Tregs that are incapable of responding to CD70 signaling perform normally in suppression assays (76).

Our finding that Tregs from CD70-intact mice become reduced when transferred into CD70-deficient recipients demonstrates that CD27–CD70 interactions in the periphery are at least as important as in the thymus for regulating circulating Treg levels. Our data resemble a previous study showing that CD28/B7 interactions regulate both thymic development and peripheral homeostasis of Tregs (93). There may be some overlap between how each pathway supports Treg levels; both CD27 and CD28 signaling promote Treg survival and self-renewal by stimulating IL-2 production by conventional T cells (82, 93). However, whereas CD28 is critical for maintaining high levels of CD25 on Tregs (93), CD27 appears to be dispensable for this purpose and even reduced CD25 expression on thymic Tregs. We have not yet determined which host cells use CD70 to sustain peripheral Treg levels, but even the limited number of CD70 constitutively expressing cells within the thymus and Peyers patches may be sufficient given that systemic Treg frequencies increase in mice transplanted with CD70-containing tumor fragments (82). Another source may be subsets of DCs and lymphocytes that transiently express CD70, perhaps as a consequence of autoimmune inflammation.

A surprising result was that the loss of CD27 signaling increased iNKT cell development and cytokine production. This contrasts with other costimulatory molecules expressed by iNKT cells (CD28, CD40L, ICOS, OX40, 4-1BB), which have been found to support iNKT cell development and/or effector functions (94–99). We ruled out that Treg and iNKT cell levels are reciprocally regulated (Supplemental Fig. 4). Thus, the most plausible explanation for our findings is that loss of CD70 alters how iNKT cells are selected in the thymus. Although iNKT cells differ from conventional αβ T cells because they are positively selected by strong rather than weak interactions with TCR-ligand complexes, studies using mice transgenically overexpressing CD28, B7, or CD1d show that intense TCR signals promote negative selection of thymic iNKT cells (67, 95). It is possible that the loss of CD70 signaling increases iNKT development by allowing a greater proportion of high-affinity iNKT cells to escape negative selection, either through lack of CD27 engagement on iNKT cells or by reducing the expression of CD1d on DP thymocytes. The iNKT1 cell subset was most affected, which is the only population that arises from stage 3 of iNKT cell development. Our observations that CD27–CD70 costimulation limits α-GalCer–induced production of IFN-γ and IL-4 contrasts with other well-known iNKT cell costimulatory pathways that augment TCR-induced cytokine production, including CD28/B7, ICOS/ICOS-L, and CD40/CD40L (94, 97, 100). Instead, they resemble a previous report that blocking the PD-1/PD-L1 coinhibitory pathway increases α-GalCer–induced production of IFN-γ and IL-4 (101), which suggests that the CD27–CD70 axis could be a coinhibitory pathway for iNKT cells. Another possibility is that CD27–CD70 costimulation facilitates the deletion of highly reactive iNKT cells in the thymus, leaving less-responsive iNKT cells to accumulate in the periphery. In either event, it appears that one function of the CD27–CD70 pathway is to attenuate peripheral iNKT cell effector responses, possibly to limit inflammation caused by the pathogenic activation of these cells.

The ability of the AT-124 CD27 agonist Ab to respectively increase and decrease Helios⁺ Treg and iNKT cell levels in vivo demonstrates that CD70 signaling modulates these immunoregulatory subsets through interacting with CD27 rather than through reverse signaling cascades. Interestingly, the effect of this Ab was much more pronounced for iNKT cells than Tregs, even though the latter express more CD27. Although short-term AT-124 treatment was able to reduce islet inflammation, the inability of long-term Ab administration to reduce T1D development or support increased Treg levels may have been due to the development of neutralizing Abs against AT-124 or the observed downregulation of CD27 expression on T cells.

In conclusion, we established that the overall effect of CD27–CD70 costimulation limits T1D in the NOD mouse, most likely by supporting the development and homeostasis of Helios⁺ Tregs through primarily a hematopoietic cell type(s). In addition, the CD27–CD70 pathway was found to be distinct from other T cell costimulatory pathways for how it plays an inhibitory role in the development and effector functions of iNKT cells. Finally, we showed that a CD27 agonist Ab reversed the effects of CD70 ablation in NOD mice, suggesting that there may be potential to alter the course of T1D by therapeutically targeting CD27.

The authors are grateful for the assistance of Dr. Yi-Guang Chen for critically reading the manuscript and providing constructive advice. We thank Dr. Michael Clare-Salzler for providing us with NOD.BDC2.5.Foxp3^eGFP mice and Celldex Therapeutics for donating the AT-124 Ab.

The authors have no financial conflicts of interest.